← All insights
Technology·Aug 17, 2026·6 min

Compute, Silicon and the Unsexy Half of Autonomy: NVIDIA, Arm and the AV Stack

Nuro lists NVIDIA and Arm among its partners. The compute layer determines what a foundation-model driver can actually do inside a vehicle.

An end-to-end foundation model has to run in real time, in a moving vehicle, within a power and thermal budget. That constraint shapes the model as much as the training data does.

Nuro publicly counts NVIDIA and Arm among its partners, alongside Uber and Lucid.

Why silicon partnerships signal seriousness

Automotive-grade compute has long qualification cycles. A company designing around named silicon partners is planning for production vehicles, not demonstrations.

It also caps unit cost, which is the difference between a pilot and a business.

The Tokyo relevance

Dense urban driving is compute-hungry: more agents, more occlusion, more interaction per kilometre. Tokyo is a stress test of the inference budget, not just the model.

Fleets that survive Tokyo economically will be deployable almost anywhere.

Where robotaxi.tokyo fits

robotaxi.tokyo is an early-stage project and a strategic domain asset — nothing is live yet. Our thesis is that Tokyo's autonomous fleets will need a neutral, consumer-facing booking, ticketing and concierge layer that sits above any single operator. The domain is open for acquisition or seed partnership: andrew.mc@nousdomains.com.

Sources & references

Primary material, official filings, and operator publications referenced while researching this article.

Strategic Asset

robotaxi.tokyo is open for acquisition or seed partnership.

The canonical URL for Tokyo's autonomous mobility layer. Talk to us before a competitor does.

Initiate inquiry →